一种新的含噪遥感图像Otsu分割算法研究

Otsu（最大类间方差）是经典的非参数、无监督、自动获取最佳阈值的最优图像分割方法。但是,在用于含噪图像的分割时,Otsu方法并不能取得理想的分割效果。针对这一问题,本文在Otsu分割方法的基础上,给出了一种新的含噪遥感图像分割算法。该算法首先用小波包对含噪图像进行全局阈值的去噪处理,然后利用局部加权回归对图像像素逐一估计去噪,得到去噪后的图像,之后采用Otsu方法对估计图像分割。仿真实验表明：该算法不仅计算量小,具有良好的抗噪能力,而且获得了较好的分割效果。

A new noisy remote image Otsu segmentation algorithm research

Otsu（the variance between the largest category） is a non-classical parameters,unsupervised,accessed to the best threshold automatic,the best image segmentation method.But when it is used in noise image,Otsu method can not provide satisfactory results,Under this situation,we try to apply new noise image segmentation method base on Otsu method.First of all,this new method apply wavelet package transformation to reduce noise,then,we apply locally weighted regresion estimation to compute each image pixel,Finally,we adopt Otsu method to segment image.The simulation experiments show that the algorithm has the small amount of computing advantages and can reach better effect of segmentation by experiment in particular Remote image segmentation.